Interferon regulatory factor 3 and type I interferons are protective in alcoholic liver injury in mice by way of crosstalk of parenchymal and myeloid cells


  • Potential conflict of interest: Nothing to report.


Alcoholic liver disease (ALD) features increased hepatic exposure to bacterial lipopolysaccharide (LPS). Toll-like receptor-4 (TLR4) recognizes LPS and activates signaling pathways depending on MyD88 or TRIF adaptors. We previously showed that MyD88 is dispensable in ALD. TLR4 induces Type I interferons (IFNs) in an MyD88-independent manner that involves interferon regulatory factor-3 (IRF3). We fed alcohol or control diets to wild-type (WT) and IRF3 knock-out (KO) mice, and to mice with selective IRF3 deficiency in liver parenchymal and bone marrow-derived cells. Whole-body IRF3-KO mice were protected from alcohol-induced liver injury, steatosis, and inflammation. In contrast to WT or bone marrow-specific IRF3-KO mice, deficiency of IRF3 only in parenchymal cells aggravated alcohol-induced liver injury, associated with increased proinflammatory cytokines, lower antiinflammatory cytokine interleukin 10 (IL-10), and lower Type I IFNs compared to WT mice. Coculture of WT primary murine hepatocytes with liver mononuclear cells (LMNC) resulted in higher LPS-induced IL-10 and IFN-β, and lower tumor necrosis factor alpha (TNF-α) levels compared to LMNC alone. Type I IFN was important because cocultures of hepatocytes with LMNC from Type I IFN receptor KO mice showed attenuated IL-10 levels compared to control cocultures from WT mice. We further identified that Type I IFNs potentiated LPS-induced IL-10 and inhibited inflammatory cytokine production in both murine macrophages and human leukocytes, indicating preserved cross-species effects. These findings suggest that liver parenchymal cells are the dominant source of Type I IFN in a TLR4/IRF3-dependent manner. Further, parenchymal cell-derived Type I IFNs increase antiinflammatory and suppress proinflammatory cytokines production by LMNC in paracrine manner. Conclusion: Our results indicate that IRF3 activation in parenchymal cells and resulting type I IFNs have protective effects in ALD by way of modulation of inflammatory functions in macrophages. These results suggest potential therapeutic targets in ALD. (HEPATOLOGY 2011;53:649-660.)

Alcoholic liver disease (ALD) is the most common drug abuse-induced liver disease and accounts for 40% of deaths from cirrhosis in the United States.1 Gut-derived lipopolysaccharide (LPS), a component of the gram-negative bacterial wall, has been proposed as a key player in the pathogenesis of ALD.2, 3 Exposure to LPS during chronic alcohol consumption results in increased production of inflammatory mediators, leading to progression of liver injury.4 Indeed, mice treated with antibiotics to eliminate gut microflora, or mice deficient in tumor necrosis factor-alpha (TNF-α) Type I receptor were protected from alcohol-induced liver injury.5, 6

Recognition of pathogen-derived molecules occurs through pattern recognition receptors such as Toll-like receptors (TLR), which are widely expressed on parenchymal and nonparenchymal cell types in the liver.7 TLR4 recognizes LPS and activates two signaling pathways by way of recruitment of adaptor molecules.8, 9 Recruitment of the common TLR adaptor, myeloid differentiation factor 88 (MyD88), leads to rapid activation of nuclear factor kappaB (NF-κB) and increased TNF-α production, whereas recruitment of TIR domain-containing adaptor inducing interferon-beta (TRIF) activates TANK-binding kinase 1/inhibitor of κB kinase epsilon (TBK1/IKKε) and interferon regulatory factor 3 (IRF3), leading to production of Type I interferons (IFNs) and delayed NF-κB activation.10-12

We previously reported that MyD88 deficiency failed to prevent alcohol-induced liver damage and inflammation, suggesting that TLR4-mediated MyD88-independent pathways are important in induction of ALD.13 The significance of MyD88-independent pathways including activation of IRF3 in ALD is yet to be evaluated.

Considering the importance of LPS-induced inflammatory activation in ALD3 and the role of MyD88-independent downstream pathways in TLR4 signaling,13 we hypothesized that IRF3 was critical in alcohol-induced liver injury. Given the differential input of parenchymal and nonparenchymal cells in the pathophysiology of ALD, we further hypothesized that IRF3 may be critical in alcoholic liver injury in a cell-specific manner. Therefore, we employed a chimeric mouse model to evaluate the effect of chronic alcohol feeding on liver damage, steatosis, and inflammation in animals with selective deficiency of IRF3 in liver parenchymal cells.

Here we demonstrate that IRF3 activation and downstream type I IFN induction in parenchymal cells have protective effects in ALD. We report that disruption of IRF3 in liver parenchymal cells decreases type I IFN production and increases liver injury due to dysregulated expression of pro- and antiinflammatory cytokines.

Abbreviations: ALD, alcoholic liver disease; BM, bone marrow; IFN, interferon; IFNAR, Type I interferon α/β receptor; IL-1β, interleukin-1 beta; IL-10, interleukin 10; IL-10R, interleukin 10 receptor; IRF3, interferon regulatory factor-3; ISG, interferon stimulated gene; KO, knock-out; LMNC, liver mononuclear cells, LPS, lipopolysaccharide; MyD88, myeloid-differentiation factor 88; NF-κB, nuclear factor κB; PBMC, peripheral blood mononuclear cells; TBK1/IKKε, TANK-binding kinase 1/inhibitor of κB kinase epsilon; TLR, toll-like receptor; TNF-α, tumor necrosis factor-alpha; TRIF, TIR domain-containing adaptor inducing interferon-beta; WT, wildtype.

Materials and Methods

Animal Studies.

All animals received proper care in agreement with animal protocols approved by the Institutional Animal Use and Care Committee of the University of Massachusetts Medical School. Six to 8-week-old, female C57Bl/6 WT, IRF3-deficient (IRF3-KO) and Type I interferon α/β receptor 1-deficient (IFNAR-KO) mice (kind gift of Jonathan Sprent, Scripps Research Institute, La Jolla, CA), all on C57Bl/6 background, were employed. Some animals were fed with the Lieber-DeCarli diet (Dyets, Bethlehem, PA) with 5% (vol/vol) ethanol (36% ethanol-derived calories) for 4 weeks; pair-fed control mice matched the alcohol-derived calories with dextran-maltose.13 Chimeric mice were generated by transplanting WT (C57Bl/6) bone marrow (BM) into irradiated, IRF3-deficient mice (IRF3-KO/WT-BM). Transplanted IRF3-deficient BM into irradiated, WT mice (WT/IRF3KO-BM), and WT mice with transplanted WT BM (WT/WT-BM) were used as controls. Serum was stored at −80°C. Livers were snap-frozen in liquid nitrogen for proteins or stored in RNAlater (Qiagen, Hilden, Germany) for RNA extraction, or fixed in 10% neutral-buffered formalin for histopathological analysis.

Bone Marrow Transplantation Protocol.

BM was collected from the long bones of 8-week-old donor male mice by flushing with a 25g needle and passed through a cell strainer to remove clumps. Female mice were irradiated (9 Gy) from a cesium irradiator (Gammacell 40, Atomic Energy of Canada), and 4 hours later transplanted with 5 million freshly isolated donor BM cells by way of a single tail vein injection. Transplanted mice were housed in microisolator cages and placed on medicated water (Sulfamethoxazole/Trimethoprim, Hi-Tech Pharmacal) until engraftment was complete 6 weeks later, as discussed.14, 15 The engraftment rate for all transplanted mice was >90%, as indicated by polymerase chain reaction (PCR) analysis of peripheral blood 6 weeks posttransplant.

Biochemical Assays.

Serum alanine aminotransferase (ALT) was determined using a kinetic method (D-Tek, Bensalem, PA). Liver triglyceride levels were assessed using the L-Type Triglyceride H kit (Wako Chemicals, VA).

Cytokine Measurement.

Mouse IL-1β enzyme-linked immunosorbent assay (ELISA) kit was purchased from R&D Systems (Minneapolis, MN), mouse and human TNF-α, IL-1β and IL-10 kits from BD Bioscience (San Jose, CA) and mouse IFN-β kit from PBL Interferon Source (Piscataway, NJ).

Protein Quantification.

Whole-cell lysates were extracted from liver and equal amounts of proteins were separated on polyacrylamide gel and transferred to a nitrocellulose membrane. Target proteins were detected by western blot and immunostaining with specific primary antibody, followed by horseradish peroxidase-labeled secondary antibody. Antibodies specific for IRF3 and IL-10 were from Santa Cruz Biotechnology (Santa Cruz, CA), antiphosphorylated IRF3 and IFN-β antibodies were from Cell Signaling Technology (Danvers, MA) and from ProSci (Poway, CA), respectively. The specific immunoreactive bands of interest were detected by chemiluminescence (Amersham, Piscataway, NJ) and quantified by densitometric analysis.

RNA Analysis.

RNA was purified using the RNeasy kit (Qiagen Sciences, MD) and on-column DNA digestion. cDNA was transcribed with the Reverse Transcription System (Promega, Madison, WI). SybrGreen-based real-time quantitative PCR was performed using the iCycler (Bio-Rad Laboratories, Hercules, CA) as described13; primer sequences are shown in Table 1.

Table 1. Real-time PCR primers
Target geneForward primer (5'→ 3')Reverse primer (5'→ 3')
18Sgta acc cgt tga acc cca ttcca tcc aat cgg tag tag cg
TNF-Acac cac cat caa gga ctc aaagg caa cct gac cac tct cc
Pro-IL-1Btct ttg aag ttg acg gac cctga gtg ata ctg cct gcc tg
IL-10ctg gac aac ata ctg cta acc gggg cat cac ttc tac cag gta a
IFN-Bagc tcc aag aaa gga cga aca tgcc ctg tag gtg agg gtt gat ct
ISG-56ggg cct tgc agg cat cac ctttcc tgc ctt ctg ggc tgc ct

Histopathological Analysis.

Liver sections were stained with hematoxylin and eosin or oil-red-O and analyzed by microscopy as described.13

Isolation of Hepatocytes and Liver Mononuclear Cells.

Anesthetized animals were perfused by way of portal vein with saline solution followed by digestion as described.13 The hepatocytes were separated by centrifugation, liver mononuclear cells (LMNCs) were purified by centrifugation in Percoll gradient.

Phenotype Analysis by Flow Cytometry.

Primary hepatocytes were washed in phosphate-buffered saline (PBS), permeabilized using the Cytofix/Cytoperm kit (BD Pharmingen, San Diego, CA), and incubated with anti-albumin (FITC) antibody (Thermo Fisher Scientific, Fremont, CA) for 30 minutes on ice. After incubation, cells were washed with permeabilization solution as indicated by the manufacturer, fixed in paraformaldehyde, and analyzed by flow cytometry.

Isolation of Human Peripheral Blood Mononuclear Cells.

Human peripheral blood mononuclear cells (PBMCs) were separated from blood of healthy volunteers by centrifugation in Ficoll gradient as described.16

In Vitro Experiments.

Primary hepatocytes and LMNCs were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 1% insulin, transferrin, selenium (ITS) solution. Primary hepatocytes were seeded in 6-well collagen-coated plates, LMNCs (106/insert) were plated in cell-culture inserts with pore diameter 0.4 μm (Becton Dickinson Labware, Bedford, MA). Before starting stimulation experiments, hepatocytes were rested for 4 hours. Subsequently, culture media was replaced and stimulation was performed as indicated in the figure legends. LPS (Sigma, St. Louis, MO) was used at 100 ng/mL. IFN-β, IL-10, and TNF-α were measured in supernatants using ELISA.

RAW264.7 macrophages were stimulated with LPS, recombinant mouse IFN-α2a (eBioscience, San Diego, CA), recombinant mouse IL-10 (PeproTech, Rocky Hill, NJ), or with antimouse IL-10 receptor antibody (Biolegend, San Diego, CA).

Human PBMCs were stimulated with LPS, recombinant human IFN-α (PBL Interferon Source), recombinant IL-10 (eBioscience), or IL-10 receptor antibody (R&D Systems).

Statistical Analysis.

Statistical significance was determined using the t-test or the nonparametric Kruskal-Wallis test using the GraphPad Prism 5.01 (La Jolla, CA). Data are shown as mean ± standard error of the mean (SEM) and were considered statistically significant at P< 0.05.


IRF3 Deficiency Protects Against Alcohol-Induced Liver Damage.

TLR4 recognizes LPS and activates two signaling pathways by utilizing the adaptor molecules MyD88 or TRIF, respectively. We showed that MyD88 is dispensable in ALD.13 In addition to induction of inflammatory cytokines by way of NF-κB, MyD88-independent activation of TLR4 triggers production of Type I IFNs, which is largely dependent on activation of intracellular pathways involving interferon regulatory factor-3 (IRF3).12 To define the importance of the MyD88-independent, IRF3-dependent signaling cascade and Type I IFNs in alcohol-induced liver injury, we fed ethanol or isocaloric control (pair feeding) diet to WT and IRF3-KO mice.

Histopathological analysis revealed that chronic alcohol feeding induced micro- and macrovesicular steatosis and inflammatory cell recruitment in ethanol-fed WT mice, suggestive of ALD (Fig. 1A). In contrast, none of the histopathological features of ALD were observed in IRF3-KO mice (Fig. 1A). Consistent with the histopathology, serum ALT levels were significantly higher in alcohol-fed WT mice, but not in the IRF3-KO mice, compared to the pair-fed controls (Fig. 1B). We also found that the expression of inflammatory cytokines TNF-α and IL-1β in the liver was significantly higher in alcohol-fed WT mice compared to pair-fed controls; this alcohol-induced proinflammatory state was absent in IRF3-KO mice (Fig. 1C,D). Alcohol feeding to WT mice triggered expression of Type I IFN stimulated gene (ISG) 56, suggesting activation of Type I IFN signaling in alcohol-induced liver injury. In contrast, alcohol feeding of IRF3KO mice failed to up-regulate ISG56 (Fig. 1E). These data suggested a role of IRF3 and/or Type I IFNs in alcohol-induced liver injury.

Figure 1.

IRF3-deficiency protects against alcohol-induced liver damage. WT and IRF3-KO mice were fed Lieber-DeCarli ethanol or control (pair-fed) diet and sacrificed after 4 weeks. Livers were fixed in formalin and stained with H&E or with Oil-red-O; magnification 200× (A). Serum ALT levels (B) were analyzed. Messenger RNA levels of liver (C) TNF-α, (D) pro-interleukin 1-beta (IL-1β), and (E) IFN stimulated gene ISG56 were analyzed by real-time PCR and normalized to 18s. Values are shown as mean ± SEM fold increase over WT pair-fed control group (9-12 mice per group). Numbers in graphs denote P-values; *P< 0.05 versus pair-fed WT mice; #P< 0.05 versus ethanol-fed WT mice.

IRF3 Deficiency in Parenchymal Cells Aggravates Alcohol-Induced Liver Damage.

The liver functions with a complex coexistence of parenchymal and nonparenchymal cells. To explore whether the protective effect of IRF3 in alcoholic liver injury is mediated by parenchymal cells or BM-derived immune cells, we generated IRF3-chimeric mice by transplanting WT BM into irradiated, IRF3-deficient mice (IRF3-KO/WT-BM mice), or by transplanting IRF3-deficient BM into irradiated WT mice (WT/IRF3KO-BM). WT mice with WT BM transplant served as controls (WT/WT-BM). As expected, WT/WT-BM mice developed ALD after 4 weeks of a Lieber-DiCarli diet, as indicated by liver steatosis, inflammatory infiltrate, and liver injury compared to pair-fed controls (Fig. 2A-C).

Figure 2.

Parenchymal cell-specific IRF3 deficiency aggravates alcohol-induced liver damage. WT mice with transplanted WT BM (WT/WT-BM), IRF3-deficient mice with transplanted WT BM (IRF3-KO/WT-BM), and WT mice with transplanted IRF3-deficient BM (WT/IRF3-KO) were fed Lieber-DeCarli ethanol or control (pair-fed) diet and sacrificed after 4 weeks. Livers were fixed in formalin and stained with H&E or with Oil-red-O; magnification 200× (A). Serum ALT levels (B) and liver triglycerides (C) were analyzed. Messenger RNA levels of liver (D) TNF-α and (F) pro-interleukin 1-beta (IL-1β) were analyzed by real-time PCR and normalized to 18s. Liver TNF-α and IL-1β levels were analyzed using ELISA (E,G). Values are shown as mean ± SEM fold increase over WT pair-fed control group (11-17 mice per group). Numbers in graphs denote P-values; *P< 0.05 versus pair-fed WT/WT-BM mice; #P< 0.05 versus ethanol-fed WT/WT-BM mice.

In sharp contrast to WT/WT-BM mice, IRF3-KO/WT-BM mice showed increased alcohol-induced liver injury, as indicated by exaggerated steatosis and inflammatory infiltrate on histology (Fig. 2A). This finding was accompanied by a significant elevation in serum ALT and in liver triglycerides compared to WT/WT-BM ethanol-fed mice (Fig. 2B,C). Further, IRF3-KO/WT-BM mice showed significantly increased expression of inflammatory cytokines TNF-α and IL-1β compared to WT/WT-BM ethanol-fed mice (Fig. 2D-G). These data suggested a protective role of IRF3 in parenchymal cells in ALD by limiting liver inflammation and injury.

Furthermore, WT/IRF3KO-BM mice showed no protection against alcohol-induced liver damage and steatosis (Fig. 2A-C) compared to WT/WT-BM mice, in spite of deficient induction of proinflammatory cytokine TNF-α (Fig. 2D,E) and IL-1β (Fig. 2G,F). These findings contrasted with the complete protection against alcohol-induced liver injury in global IRF3-KO mice (Fig. 1), and suggested that both parenchymal cell-specific and myeloid-specific IRF3 is required for the pathogenesis of alcohol-induced liver damage. More important, they indicated a protective role of parenchymal cell-specific IRF3 in alcohol-induced liver damage.

Deficiency of IRF3 in Parenchymal Cells Is Associated with Decreased Type I IFN and IL-10 Induction.

Activation of IRF-3 leads to preferential induction of IFN-β.17 We identified that, in contrast to WT and to WT/IRF3KO-BM mice, IRF3-KO/WT-BM mice showed a significantly decreased expression of IFN-β (Fig. 3A) and of interferon-inducible gene ISG-56 (Fig. 3B). This finding indicated that aggravated liver injury in IRF3-KO/WT-BM mice is associated with deficiency in IRF3-dependent type I IFNs induction and signaling and suggested possible involvement of IRF3- and Type I IFN-dependent antiinflammatory factors in alcohol-induced liver injury. We thus analyzed the expression of the Type I IFN-dependent gene, IL-10, which is a major antiinflammatory cytokine.18 The extent of IL-10 increase was significantly higher in the liver of WT/WT-BM and WT/IRF3KO-BM mice compared to IRF3-KO/WT-BM mice after alcohol feeding (Fig. 3C,D); the latter showed a significantly lower baseline IL-10 expression compared to controls (Fig. 3C,D). Collectively, these findings suggested that parenchymal cell-specific IRF3 is required for expression of IFN-β and IL-10 in alcohol-induced liver injury.

Figure 3.

Parenchymal cell-specific IRF3 deficiency is associated with decreased Type I IFN and IL-10 induction. WT mice with transplanted WT BM (WT/WT-BM), IRF3-deficient mice with transplanted WT BM (IRF3-KO/WT-BM) and WT mice with transplanted IRF3-deficient BM (WT/IRF3-KO) were fed Lieber-DeCarli ethanol or control (pair-fed) diet and sacrificed after 4 weeks. Messenger RNA levels of (A) liver interferon β (IFN-β), interferon stimulated gene 56 (ISG-56), and (C) interleukin 10 (IL-10) were analyzed by real-time PCR and normalized to 18s. Liver IL-10 protein levels were analyzed using immunoblot analysis (D). Values are shown as mean ± SEM fold increase over WT pair-fed control group (11-17 mice per group). Numbers in graphs denote P-values. *P< 0.05 versus pair-fed WT/WT-BM mice; #P< 0.05 versus ethanol-fed WT/WT-BM mice.

Hepatocytes Produce Type I IFN in an IRF3- Dependent Manner.

Our findings suggested that expression of liver IL-10 is linked to activation of IRF3 in parenchymal cells. To confirm that the parenchymal cell-specific role of IRF3 is attributable to hepatocytes, we isolated primary hepatocytes from WT mice and observed >97% purity of hepatocyte isolates (Fig. 4A). Next, we stimulated primary WT hepatocytes with LPS ex vivo and observed induction of IRF3 phosphorylation, which was matched by induction of IFN-β (Fig. 4B). Although statistically significant induction of IFN-β mRNA and protein was observed in WT hepatocytes, no IFN-β induction occurred in hepatocytes deficient in IRF3 at the mRNA or protein levels (Fig. 4C,D). These in vitro observations suggested that hepatocytes are a major source of IFN-β in ethanol/LPS-induced liver injury.

Figure 4.

Parenchymal cells induce Type I IFN in an IRF3-dependent manner. Primary murine hepatocytes were permeabilized, stained with anti-albumin antibody, and analyzed by flow cytometry (A). WT primary hepatocytes were stimulated with 100 ng/mL LPS for indicated timepoints and levels of phosphorylated IRF3 (p-IRF3), total IRF3, IFN-β, and beta tubulin and beta actin were analyzed using immunoblotting (B). WT and IRF3-deficient (IRF3-KO) primary murine hepatocytes were stimulated LPS, and messenger RNA levels of interferon β (IFN-β) were analyzed by real-time PCR and normalized to 18s (C). IFN-β protein levels in supernatant were analyzed using ELISA (D). Values are shown as mean ± SEM (5 mice per group). Numbers in graphs denote P-values; *P< 0.05 versus nonstimulated WT hepatocytes; #P< 0.05 versus LPS-stimulated WT hepatocytes.

Hepatocyte-Specific Type I IFNs Modulate Cytokine Production in Liver Mononuclear Cells.

We further employed cocultures of hepatocytes and LMNCs to dissect the regulatory loops involved in Type I IFN/IL-10 production. Control unstimulated and LPS-stimulated WT hepatocytes produced significantly more IFN-β than LMNCs (Fig. 5A,B, groups 1 and 3). On the contrary, IL-10 was produced mainly by LMNCs (Fig. 5C,D, group 3), which supports the data that Kupffer cells stimulated with LPS produce IL-10.19, 20 Importantly, LMNCs coculture with primary hepatocytes resulted in increased IL-10 production compared to either cell types alone, which was further significantly increased upon stimulation with LPS (Fig. 5C,D, groups 1, 3, 5). The induction of IL-10 in hepatocyte/LMNC coculture exceeded a merely additive contribution of both cell types to the secretion of IL-10, suggesting that hepatocyte-derived IFN-β facilitates the production of IL-10 in LPS-challenged immune cells in the liver. In contrast, we observed significantly lower induction of IL-10 in LPS-stimulated cocultures of hepatocytes and LMNCs from IRF3-KO or IFNAR-KO mice (Fig. 5C,D, groups 6, 7), or in cocultures of WT hepatocytes with IFNAR-deficient LMNCs (Fig. 5D, group 8) compared with cocultures of WT hepatocytes and WT LMNCs (Fig. 5C,D, group 5). These findings supported our hypothesis that enhancement of LPS-induced IL-10 expression in LMNCs is dependent on production of Type I IFN in parenchymal cells.

Figure 5.

Parenchymal cell-derived Type I IFNs up-regulate IL-10 and down-regulate TNF-α in liver mononuclear cells stimulated with LPS. Primary hepatocytes and LMNCs were isolated from WT, IRF3-deficient mice (A,C,E) or Type I interferon α/β receptor 1-deficient (IFNAR-KO) mice (B,D,F) and stimulated with 100 ng/mL LPS in transwell cell-culture systems as indicated. Protein levels of (A,B) IFN-β, (C,D) IL-10, and (E,F) tumor necrosis factor α (TNF-α) in cell-free supernatants were analyzed using specific ELISA. Values are shown as mean ± SEM (5 mice per group). Numbers in graphs denote P-values. *P< 0.05 versus nonstimulated hepatocytes of the respective genotype; #P< 0.05 versus LPS-stimulated hepatocytes of the respective genotype.

Given the tight control of the pro- and antiinflammatory balance in the liver, we further asked whether Type I IFN-dependent IL-10 production may affect the level of TNF-α in liver immune cells. We identified that TNF-α production by WT LMNCs was significantly down-regulated upon their coculture with WT hepatocytes (Fig. 5E,F, groups 3, 5); such an effect was absent in LMNCs of IFNAR-KO mice (Fig. 5F, groups 4, 6, 8). We observed that stimulation with LPS induced TNF-α also in IRF3-deficient LMNCs (Fig. 5E, group 4), whereas no induction was observed in IRF3-deficient livers of alcohol-fed mice (Figs. 1C, 2D,E). This finding suggests a context-specific role for IRF3: whereas IRF3 is crucial for the synergism of ethanol and LPS in induction of TNF-α from monocytes/macrophages, other transcription factors likely induce TNF-α in monocytes/macrophages that are stimulated only with LPS.21

To further evaluate if regulation of IL-10 by Type I IFNs is preserved across species, we stimulated RAW264.7 murine macrophages or human PBMCs with LPS and Type I IFN and identified a significant increase of IL-10 in the presence of Type I IFN compared to stimulation with LPS alone in both species (Fig. 6A,B). We found that expression of TNF-α was significantly decreased in RAW264.7 murine macrophages stimulated with LPS in the presence of recombinant IL-10, whereas neutralization of the IL-10 receptor (IL-10R) with anti-IL10R antibody significantly up-regulated TNF-α (Fig. 6C). We also observed a dose-dependent inhibition of TNF-α and IL-1β by IL-10 in human PBMCs (Fig. 6D). Further, a 50% inhibitory concentration (IC50) of IL-10 caused a significant inhibition of LPS-triggered TNF-α and IL-1β (Fig. 6D-F), whereas inhibition of IL-10 receptors using IL-10R antibody significantly up-regulated secretion of inflammatory cytokines in human PBMCs (Fig. 6E,F). These data confirmed that Type I IFN potentiates the LPS-induced IL-10 both in human and mouse systems.

Figure 6.

Type I IFN-dependent induction of IL-10 modulates inflammatory cytokines in mononuclear cells. Murine RAW264.7 macrophages were stimulated with 100 ng/mL LPS, 1000 IU/mL murine recombinant IFNα2a, 10 ng/mL recombinant murine IL-10, and 1 μg/mL antimouse IL10 receptor antibody (anti IL-10R Ab). Human PBMCs (n = 4) were stimulated with 10 ng/mL LPS, 1000 IU/mL human recombinant IFNα2, 10-2000 pg/mL recombinant human IL-10, and 1 μg/mL anti IL-10R Ab. Protein levels of (A,B) interleukin 10 (IL-10), (C-E) TNF-α, and (D,F) IL-1β were analyzed using ELISA. Values are shown as mean ± SEM. Numbers in graphs denote P-values; *, #, §, †P< 0.05 versus respective control group.

Taken together, these data demonstrate that parenchymal cell-derived Type I IFNs up-regulate IL-10 and down-regulate inflammatory cytokines in nonparenchymal cells in the liver. More important, they outline the paradigm of intercellular cooperation and regulation in the liver, where hepatocytes control the inflammatory potential of immune cells.


Chronic consumption of ethanol is tightly linked to liver inflammation and steatosis in human disease as well as in experimental models. Whereas activation of TLR4-dependent pathways by gut-derived LPS and induction of inflammatory cytokines has been traditionally attributed to BM-derived Kupffer cells,22 the role of crosstalk between parenchymal and nonparenchymal (BM-derived immune cells) in ALD remains elusive. Here we demonstrate that liver response to insults is a multistep process: IRF3 in parenchymal cells drives Type I IFN induction in the liver and parenchymal cell-derived Type I IFN leads to a modulation of inflammatory cytokines in nonparenchymal BM-derived cells (Fig. 7). Our novel findings outline a link between parenchymal and liver immune cells in modulation of innate immune signaling in ALD.

Figure 7.

Proposed mechanism of parenchymal cell-mediated control of inflammatory responses in alcohol-induced liver injury. Chronic alcohol consumption results in increased exposure of the liver to the gut-derived LPS. LPS is recognized by way of TLR-4 on nonparenchymal and parenchymal cells. In nonparenchymal cells, LPS increases production of inflammatory cytokines by way of multiple transcription factors, including IRF3. In parenchymal cells, LPS induces Type I IFNs in an IRF3-dependent manner. In turn, hepatocyte-derived Type I IFNs enhance IL-10 and down-regulate TNF-α production in nonparenchymal cells, thus regulating the balance between inflammatory and antiinflammatory cytokines in the liver.

BM-derived cells are considered the main targets of pathogen-derived products in the liver due to their strategic position to encounter pathogens in the portal system and broad-range expression of TLRs. Our study provides novel lines of evidence that parenchymal cells are the main producers of Type I IFNs in response to alcohol/LPS exposure, and that IRF3 is a dominant signaling molecule inducing Type I IFN in alcoholic liver disease.

First, chimeric mice containing IRF3-deficient liver parenchymal cells and WT BM-derived cells show a similar reduction in baseline and ethanol-induced expression of Type I IFNs as mice with global IRF3 deficiency. Second, no decrease in liver expression of Type I IFNs was observed in mice with selective deficiency of IRF3 in BM-derived cells. Third, ex vivo stimulation of WT primary mouse hepatocyte isolates with LPS resulted in phosphorylation of IRF3 and in a significant up-regulation of Type I IFNs, in contrast to hepatocyte isolates from IRF3KO mice that failed to induce Type I IFNs. In addition, phenotypic analysis of hepatocyte isolates employed in our study indicated that the IRF3-dependent Type I IFN induction indeed originates from hepatocytes, whereas the role of other cell types remains negligible.

Our study defines induction of Type I IFNs by way of IRF3 in hepatocytes and down-regulation of inflammatory cytokines in BM-derived cells as two complementary, yet independent mechanisms by which TLR4 controls the extent of alcohol-induced liver inflammation and injury. Kupffer cells stimulated by way of TLR4 are a main source of inflammatory cytokines in the liver and promote tissue inflammation, injury, and fibrosis.22 Thus, TLR4 seems to activate IRF3 in both parenchymal and nonparenchymal liver cells: here we demonstrate that, whereas the signaling pathways are shared, we observed a cell-specific response to LPS, with a distinct outcome.

Studies by Zhao et al.21 suggested that IRF3, activated by TLR4/TRIF and ethanol, induces inflammatory cytokines in macrophages, thereby playing a proinflammatory role. We observed no induction of inflammatory cytokines in mice with BM-specific deficiency of IRF3; however, our novel data show that this effect was not sufficient to prevent alcohol-induced liver injury. These findings suggest that both myeloid and parenchymal cell-specific IRF3 contribute to ALD, i.e., that the solo contribution IRF3 in BM-derived cells is not sufficient for the development of ALD. The type of signal in IRF3 deficient BM-derived cells that improves ALD in the global IRF3 knockouts, and the reason why this signal requires the absence of IRF3 in parenchymal cells remains to be further investigated.

Recently, Klein et al.23 and Kennedy and Abkowitz24 reported that in chimeric mice two populations of liver macrophages coexist: radioresistant macrophages that show tolerogenic properties, and radiosensitive macrophages that are immunogenic; the latter macrophages are rapidly replaced by BM transplantation and expected to be the dominant subtype that participates in the immunoinflammatory reactions in the liver posttransplant. Thus, the lack of protection from ALD in mice with BM-specific deficiency of IRF3 could be attributable, at least in part, to an incomplete substitution of the host with BM-derived cells and/or differential host/donor macrophage ratio,24 either of which could influence the nature of the response to LPS and/or alcohol.

The differential role of IRF3 in ALD seems to be dominated by its parenchymal cell-specific protective effect. Our data demonstrate that IRF3 in parenchymal cells dampens TLR4-induced inflammatory response by an indirect (paracrine) mechanism mediated by Type I IFNs. The importance of this cell-specific activation of IRF3 and Type I IFNs is emphasized by our finding that aggravated liver inflammation and injury was observed in mice chimeras lacking IRF3 in parenchymal cells, and was further associated with a significantly decreased expression of IL-10, a major antiinflammatory cytokine, in the liver.

Our finding of Type I IFN-dependent induction of the antiinflammatory state in the liver is supported by the fact that the IL-10 promoter contains a Type I IFN-dependent responsive element,25 which makes this cytokine a Type I IFN-dependent antiinflammatory mediator. We found that, ex vivo, LPS-stimulated liver mononuclear cells synthesized significantly more IL-10 when cocultured with primary hepatocytes that produced significant amounts of Type I IFNs. This synergism was completely absent in cocultures of WT hepatocytes with IFNAR1-deficient LMNCs, but only partially abolished in cocultures containing LMNCs that lacked IRF3, suggesting that it is the parenchymal cell-derived Type I IFN that acts synergistically with LPS on LMNCs to produce IL-10, rather than IRF3 in LMNCs per se. The existence of a hepatocyte/immune cell regulation loop is further supported by our finding that the facilitation of LPS-induced production of IL-10 by hepatocyte-specific Type I IFNs in liver mononuclear cells was abrogated in cells lacking Type I IFN receptor. Furthermore, our data show that administration of IL-10 to mouse macrophages or human PBMCs stimulated with LPS significantly suppresses inflammatory cytokines, and therefore support the critical role of IL-10 in determining the pro- and antiinflammatory balance in the pathogenesis of ALD.19, 26 Taken together, these findings demonstrate that full expression of antiinflammatory factors in BM-derived cells is dependent on Type I IFN signaling from parenchymal cells, which is regulated by IRF3.

TLRs fulfill a variety of functions in the liver, and inhibition of TLR4 signaling may alter biological processes related to liver inflammation, injury, and fibrosis.27-30 TLR4 also promotes disease progression in alcoholic and nonalcoholic steatohepatitis,13, 31 primary sclerosing cholangitis,32 and ischemia-reperfusion injury,33 and therefore represents a potential therapeutic target. Indeed, use of probiotics, antifibrotics, or antiinflammatory agents are proposed as potential therapeutic options for these diseases.34 However, excessive TLR signaling triggers not only harmful responses, but also beneficial responses, such as clearance of microorganisms,35 tissue regeneration,36 and, as we demonstrate in this study, indirect induction of antiinflammatory loop by way of induction of protective Type I IFNs in hepatocytes and antiinflammatory IL-10 in macrophages in an IRF3-dependent manner. Therefore, it seems plausible that fine tuning, in contrast to approaches that would completely abrogate TLR signaling, may have a future in efforts to translate TLR pathophysiology into clinical practice in human liver diseases.